Brookfield, NextEra to Build $100 Billion AI Hub on Cold War Nuclear Site

The Paducah American Energy Hub is a planned AI and high-performance-computing campus in Western Kentucky. The U.S. Department of Energy announced the private investment as one of Kentucky’s largest economic development projects. Its purpose is to combine large-scale computing with dedicated energy infrastructure.

Brookfield Asset Management will build the 1.8-gigawatt AI and HPC data center campus. NextEra Energy will develop 2 gigawatts of natural gas-fired generation and 2.6 gigawatts of battery energy storage. Together, these facilities are intended to supply the campus without depending entirely on ordinary utility service.

The project is targeted for phased completion by 2032. Its design also allows excess electricity to flow back to the regional grid. That could help the campus support local reliability while avoiding higher utility bills for nearby families and businesses. A closed-loop cooling system is planned to reduce outside water use.

The planned Paducah project is enormous by both economic and infrastructure standards. The announced private investment totals $100 billion. Brookfield’s campus will have 1.8 gigawatts of AI and high-performance-computing capacity, while NextEra will build 2 gigawatts of natural gas generation and 2.6 gigawatts of battery storage.

Construction is expected to create roughly 8,000 jobs. Once operating, the hub is expected to support about 600 permanent operations jobs. The combined generation and storage figures represent 4.6 gigawatts of planned energy infrastructure, separate from the data center’s computing capacity.

The buildout is phased and targeted for 2032. The project is designed to supply its campus with dedicated power and send excess electricity to the regional grid. Its cooling system will recirculate water in a closed loop. These features aim to manage the hub’s large energy demand without raising local utility bills or heavily consuming outside water.

The former Paducah Gaseous Diffusion Plant was a Cold War-era uranium-enrichment facility. That history made the location suitable for another very large industrial project. The proposed AI campus needs substantial electricity, cooling, and supporting infrastructure, so an established industrial site offers a logical setting.

The earlier plant was designed to draw up to 30 million gallons of water each day from the Ohio River. The new project instead plans a modern closed-loop cooling system that continuously recirculates water. It also includes dedicated power infrastructure rather than relying only on ordinary local utility supplies.

The article presents the hub as far more resource-efficient than its predecessor. That comparison matters because AI and high-performance computing can create heavy demands for electricity and cooling. Reusing an industrial location may support the project’s scale, but the article does not provide a detailed account of which existing buildings or utility assets will be reused.

An AI or high-performance-computing data center is a specialized facility filled with powerful computers, networking equipment, and systems that store and move data. It provides computing capacity for demanding tasks such as training AI models and running them for users. The Paducah plan centers on this kind of campus.

AI workloads use many advanced chips at once. Those chips perform enormous numbers of calculations, so the computers draw substantial electricity. Nearly all that electrical energy ultimately becomes heat. Cooling equipment must remove that heat continuously to keep the hardware operating reliably. The article specifically connects the proposed campus with a closed-loop cooling system.

The scale creates a major infrastructure challenge. A large campus needs dependable power and a steady way to manage heat without excessive water use. Paducah’s planned 1.8-GW campus is paired with generation, batteries, and recirculating cooling. The article’s broader examples show why electricity prices, permits, financing, and environmental concerns can affect AI construction plans.

The hub’s dedicated power supply is intended to separate its heavy demand from ordinary local utility service. NextEra plans 2 GW of gas-fired generation and 2.6 GW of battery storage. Together, they could help match supply with the campus’s changing computing needs and provide power when generation is unavailable.

The project is also designed to send excess electricity back to the regional grid. The article says this approach is intended to avoid raising utility bills for local families and businesses. Batteries can store electricity and release it later, although the article does not specify their operating schedule or market rules.

Cooling is another important effect. The planned closed-loop system will continuously recirculate water and minimize external consumption. That contrasts with the former plant’s design capacity to draw up to 30 million gallons daily from the Ohio River. The project could therefore reduce new pressure on local water supplies while adding generation and storage to the regional system.

Large AI projects require unusual amounts of money, electricity, land, and infrastructure. That concentration creates risks for lenders and investors, while regulators must review energy and land-use plans. Communities have also raised environmental concerns about data centers across the United States, Europe, and Asia.

Stargate illustrates the financial problem. JPMorgan Chase struggled to syndicate $38 billion in construction debt because investors disliked exposure to single-tenant AI facilities. Buyers demanded higher returns, forcing banks to sell debt at a discount. Axiom also paused Stargate UK because of high industrial electricity prices and regulatory uncertainty.

Project Jupiter shows the permitting problem. Oracle stalled its 1,400-acre New Mexico campus after failing to secure state and federal permits for a natural-gas plant. Oracle proposed solid oxide fuel cells instead, but the article calls the technology experimental at that scale. These setbacks suggest future projects will need stronger financing, permits, power plans, and community support.

Data centers provide computing as a service. Electricity powers servers, AI accelerators, storage systems, and networking equipment. Computer chips perform the calculations needed to train AI models or run them in response to requests. The facility’s software and hardware work together so customers can access computing without owning the entire campus.

During training, chips repeatedly process large amounts of data and adjust a model. During operation, they calculate answers from a trained model. Networks carry data into the facility and results back to users. Storage systems hold models and data. Cooling systems remove the heat produced by the continuously operating equipment.

The article’s Paducah plan assembles all these needs at unusual scale. Brookfield will build a 1.8-GW AI and HPC campus, while NextEra supplies generation and batteries. The closed-loop cooling system supports the computing equipment while limiting outside water use. The article does not describe specific chip models or software, so those details are general industry context.